The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local utility scripts (doctor.sh, graph.sh, build-index.sh) located within the ../woostack-init/scripts/ directory. It also uses standard git commands (git ls-files, git log) to identify files and retrieve history for context.
[DATA_EXPOSURE]: The tool reads repository content, including internal memory notes (.woostack/memory/), documentation files, and git logs to analyze project knowledge. All operations are restricted to the local working directory.
[PROMPT_INJECTION]: The skill accepts an optional instructions argument and processes repository content (notes and docs) which could theoretically contain instructions designed to influence the agent's behavior. This risk is mitigated by a 'Hard Review Gate' in Phase 3, which requires explicit, unambiguous user approval of a full changeset before any modifications are applied.
[SAFE]: The skill adheres to 'local-only' principles, ensuring memory changes are never automatically committed or merged, and documentation edits are left in the working tree for manual review.

woostack-dream